arXiv subjects

Luis G. León-Vega

Publications and source records attributed to Luis G. León-Vega.

3 recordsLinked to original sources

arXiv2025

A Quantitative Evaluation of Approximate Softmax Functions for Deep Neural Networks

The softmax function is a widely used activation function in the output layers of neural networks, responsible for converting raw scores into class probabilities while introducing essential non-linearity. Implementing Softmax efficiently poses challenges on low-end FPGAs due to limited hardware resources and the computational complexity of exponential and division operations. This work evaluates approximate computing techniques for softmax acceleration using Taylor series and interpolation methods using Look-Up Tables (LUTs). These approximations aim to reduce execution time and resource consumption while maintaining acceptable levels of numerical precision. Our findings show that quadratic interpolation with LUTs yields the lowest numerical error. In contrast, Taylor-based approximations offer significantly better performance in terms of execution time and resource efficiency due to their computational simplicity. When applied to real-world deep learning models such as LeNet-5 and MobileNet v2, the first- and second-order Taylor approximations provided substantial trade-offs between accuracy and resource savings, achieving up to 0.2% accuracy degradation and 14% resource reduction compared to exact implementations. These results highlight the effectiveness of approximate Softmax designs on resource-constrained FPGAs and lay the groundwork for their integration into larger models, including large language models (LLMs).

cs.AR↗

arXiv2024

EfiMon: A Process Analyser for Granular Power Consumption Prediction

High-performance computing (HPC) and supercomputing are critical in Artificial Intelligence (AI) research, development, and deployment. The extensive use of supercomputers for training complex AI models, which can take from days to months, raises significant concerns about energy consumption and carbon emissions. Traditional methods for estimating the energy consumption of HPC workloads rely on metering reports from computing nodes power supply units, assuming exclusive use of the entire node. This assumption is increasingly untenable with the advent of next-generation supercomputers that share resources to accelerate workloads, as seen in initiatives like Acceleration as a Service (XaaS) and cloud computing. This paper introduces EfiMon, an agnostic and non-invasive tool designed to extract detailed information about process execution, including instructions executed within specific time windows and CPU and RAM usage. Additionally, it captures comprehensive system metrics, such as power consumption reported by CPU sockets and PSUs. This data enables the development of prediction models to estimate the energy consumption of individual processes without requiring isolation. Using a regression-based mathematical model, our tool is able to estimate single processes' power consumption in isolated and shared resource environments. In shared scenarios, the model demonstrates robust performance, deviating by a maximum of 2.2% on Intel-based machines and 4.4% on AMD systems compared to non-shared cases. This significant accuracy showcases EfiMon's potential for enhancing energy accounting in supercomputing, contributing to more efficient and energy-aware optimisation strategies in HPC.

cs.DC↗

arXiv2024

A Comprehensive Analysis of Process Energy Consumption on Multi-Socket Systems with GPUs

Robustly estimating energy consumption in High-Performance Computing (HPC) is essential for assessing the energy footprint of modern workloads, particularly in fields such as Artificial Intelligence (AI) research, development, and deployment. The extensive use of supercomputers for AI training has heightened concerns about energy consumption and carbon emissions. Existing energy estimation tools often assume exclusive use of computing nodes, a premise that becomes problematic with the advent of supercomputers integrating microservices, as seen in initiatives like Acceleration as a Service (XaaS) and cloud computing. This work investigates the impact of executed instructions on overall power consumption, providing insights into the comprehensive behaviour of HPC systems. We introduce two novel mathematical models to estimate a process's energy consumption based on the total node energy, process usage, and a normalised vector of the probability distribution of instruction types for CPU and GPU processes. Our approach enables energy accounting for specific processes without the need for isolation. Our models demonstrate high accuracy, predicting CPU power consumption with a mere 1.9% error. For GPU predictions, the models achieve a central relative error of 9.7%, showing a clear tendency to fit the test data accurately. These results pave the way for new tools to measure and account for energy consumption in shared supercomputing environments.

cs.DC↗